The reaction of the WSE to U.S.
employment news announcements
Managerial Economics 14, 165-176
Milena Suliga*, Tomasz Wojtowicz*
The reaction o f the WSE
to U.S. employment news announcements
1. Introduction
From the financial and economic literature it follows that macroeconomic news plays important role in asset pricing as it carries information about condi tions of the economy and possible future government actions. Previous studies show that not only domestic news is significant for investors. Sometimes informa tion about other economies is even more important, particularly when it is the U.S. economy. Such conclusion is obtained by e.g. Nikkinen and Sahlström [12], Będowska-Sójka [3], Harju and Hussain [11] and Gurgul et al. [10]. Their studies are performed on the basis o f daily data as well as intraday data.
Using daily data Nikkinen and Sahlström [12] study the impact o f macroeco nomic announcements on German and Finnish equity markets. The data set con sidered in the paper consist o f monthly reports from Germany, Finland and the U.S. involving three prominent macroeconomic indicators, namely Consumer Price Index, Producer Price Index and Unemployment Rate. Results o f the study show that in the period from January 1996 to December 1999 neither the German nor the Finnish market is sensitive to domestic macroeconomic news but both o f them react to U.S. announcements. UR and PPI influence the implied volatility on both stock markets while CPI reports have an impact on uncertainty on the Finnish market only.
The issue whether the U.S. economy is an important source of information to European investors is also studied by Harju and Hussain [11]. They use a wide set o f U.S. macroeconomic releases from September 2000 to March 2006. Harju
* AGH University of Science and Technology, Faculty of Management, Department of Application of Mathematics in Economics, e-mail: msuliga@zarz.agh.edu.pl, twojtow@agh.edu.pl
Financial support for this paper from the National Science Centre o f Poland (Research Grant DEC- 2012/05/B/HS4/00810) is gratefully acknowledged.
and Hussain study their impact on high-frequency returns of four European indi ces: CAC40, DAX30, FTSE100 and SMI. For all o f these stock markets Harju and Hussain find that intraday returns as well as volatilities react immediately to U.S. macroeconomic surprises.
In the similar manner Będowska-Sójka [3] studies the reaction of WIG20, DAX and CAC40 indices to announcements of eight types o f US macroeconom ic indicators, published between November 2007 and May 2009- She analyzes the impact o f individual types of announcements as well as market reaction to “good” and “bad” news. Będowska-Sójka shows that U.S. macroeconomic news changes both returns and volatilities. Unemployment Rate announcements, on the average, decrease returns and increase volatilities o f all three European indi ces under study.
The impact o f announcements about CPI, UR and Industrial Production in the U.S. on the Polish stock market is also studied by Gurgul et al. [10]. The au thors consider daily returns o f WIG20 from February 2004 to December 2011. Using the event study analysis Gurgul et.al. show that investors significantly re spond to CPI and IP announcements while UR releases seem to have no impact on the Warsaw Stock Exchange.
The abovementioned papers prove that releases o f U.S. macroeconomic indicators do really impact developed as well as emerging European equity markets. This point o f view is reinforced by Andersen et al. [2] who investigate the impact o f U.S. fundamentals releases on stock, bond and exchange mar kets in the U.S., the UK and Germany. Their data set consists o f 22 U.S. mac roeconomic indicators and 5-minute returns from July 1, 1998 to December 31, 2002. Andersen et al. not only confirm the significant impact of U.S. an nouncements on the conditional means of European stock markets but also show that the strength and direction o f the impact depends on the state of the U.S. economy i.e. traditionally bad information have a negative impact on European markets during contractions but the effect is positive while expan sions. This is in line with the previous results of Boyd et al. [4] who study the impact of the U.S. Unemployment Rate announcements on S&P500 stock index in the period from February 1948 to December 2000. Boyd et al. discover that during contractions average stock returns are positive on the day when “good” news is announced and negative on the day o f “bad” news. On the other hand, during expansions the situation is quite different - averages are positive on the day of the announcement regardless o f the information contained in the release. Moreover, the market reaction to an unexpected rise in unemployment is positive in expansions and negative in contractions. Following Campbell and Mei [5] Boyd et al. explain that during contractions UR releases carry important information about interest rates, while during contractions in UR
announcements dominates information about equity risk premium and future corporate dividends.
From the abovementioned papers it follows that the state o f the labor mar ket is one of the most frequently examined macroeconomic issue. In fact, the Employment Report published monthly by the U.S. Bureau o f Labor Statistics is one o f the most significant U.S. macroeconomic announcements ([ 1], [2], [6] ). However, in the majority o f papers the state of the U.S. labor market is charac terized by the Unemployment Rate which is only one o f the indicators in the Report. Other important indicators are e.g. Average Hourly Earnings, Average Workweek, Nonfarm Payrolls, Manufacturing Payrolls Change and Private Payrolls Change. All of these indicators are released simultaneously and joint ly affect price formation processes. Thus, the reaction o f the stock markets on the Employment Report depends on the level o f each o f these indicators. Studies that focus only on the value o f the Unemployment Rate ignore the fact that it describes one side o f the employment situation in the U.S. and other indicators provide additional information. In particular, there could be an in dicator that is more informative and have greater impact on investor behavior than UR.
In this paper we analyze the impact of four indicators published in the Employment Report on the Warsaw Stock Exchange. On the basis of high-fre quency WIG returns from January 2004 to November 2012 we evaluate the sig nificance of each of these indicators in price formation processes on the WSE. To indicate the most important indicator we also study the joint impact o f some pairs of indicators from the Employment Report. The analysis in the paper esti mates the duration and direction of the impact of U.S. employment news on the Polish stock market.
The structure o f the paper is as follows. Data and methodology applied in the study are described in detail in the next section. Empirical results of the event study analysis are presented in Section 3- The final section concludes the paper.
2. Data and methodology
From the group o f indicators that are published monthly in Employment Reports of the Bureau of Labor Statistics we choose four, that are commonly provided by information agencies and business and financial services. These are Unemployment Rate, Nonfarm Payrolls, Average Hourly Earnings and Average Workweek. Each o f them describes various aspects of the employment situ ation and its changes can lead to different investor reaction. Decrease in the
Unemployment Rate is usually good news. However, too low UR can be inter preted as a symptom of labor shortage leading to wage growth. The level of the economic activity is reflected by changes in Nonfarm Payrolls, which measure monthly changes in the number of employed outside the farming industry. The more workplaces, the faster economic growth. Nevertheless, for the economy in the mature phase of expansion rapid growth of the employment can lead to inflationary pressures. Average Workweek expresses the weekly amount of work ing hours. Rising Average Workweek causes the increase in employment and sup ports production gains while decreasing Average Workweek could lead to a de cline in employment. Wage pressure is represented by increase in Average Hourly Earnings. It is commonly believed that the upward tendency of this indicator can yield higher inflation. On the other hand, product prices do not increase if the wage growth are caused by the growth of productivity because the unit labor costs are unchanged.
In this paper we examine the reaction of stock prices on the WSE to unex pected news about each o f the abovementioned indicators. Dates and hours of announcements as well as released values and their forecasts come from www. bloomberg.com and cover the period from January 2004 to November 2012.
Employment Reports are released on the first Friday o f the month at 8:30 EST (Eastern Standard Time) i.e. usually at 14:30 CET (Central European Time). Due to differences in introducing the Daylight Saving Time in the U.S. and Europe one announcement (November 2, 2012) arrived the WSE at 13-30 CET. From the data sample we exclude 5 events when there was no trading on the WSE in the day o f the announcement. It gives the total number o f 102 report releases.
For each indicator under study we compared the announced value with its previous forecast and thus we divided all announcements into three clusters: “below consensus”, “in line with consensus” and “above consensus”. Unexpected news are connected with events in the first and third clusters. To analyze the impact of these unexpected news on the WSE we applied the event study. More precisely, we applied nonparametric rank test proposed by Corrado and Zivney [9] to test the significance of abnormal 5-minute returns o f WIG.
It is well established in financial literature that intraday volatility of high- frequency returns is higher at the beginning and at the end o f a trading session. In this study we consider the pre-event window of the length 48 returns (four hours) and the event window o f the length o f 35 minutes. The event window consists o f seven 5-minute returns: two returns before the announcement, the event return and four returns after the release. The Employment Report is usu ally released at 14.30 CET, thus the earliest return in pre-event window is from
10.25 CET and the latest return in event window is from 14.55 CET. When the employment report is released at 13-30 CET then we use 5-minute returns from 9-25 to 13-55 CET i.e. 25 minutes after opening o f the trading session and more than three hours before closing. Such pre-event and event windows do not in clude returns from the beginning nor the end o f the trading session when in creased volatility is observed.
If we denote the moment of the news release by t = 0 then the pre-event window are t = -5 1 , ..., - 3 while t = -2 , ..., 4 are event window. Abnormal re turns AR.t for г'-th event are defined as differences between actual returns and the mean from the pre-event window. For each event abnormal returns are standardized:
is abnormal returns standard deviation in the pre-event window. Increased vola tility of abnormal returns is frequently observed in the event window ([7], [8]). It is also visible in the case o f the Employment Report announcements.
Figure 1 presents cross-sectional standard deviations o f abnormal returns around the time of the announcement, i.e. at the end of the pre-event window (t = -9 , ..., -3 ) and in the event window (t = -2 , ..., 4). It shows that just after the event, for t > 3, cross-sectional volatility of abnormal returns increases. To control for this event-induced shift in the cross-sectional variance we adjust stan dardized abnormal returns:
is cross-sectional standard deviation and N is the number of events in a cluster. We tested the significance of abnormal returns for each t in the event window sepa rately thus for each iQ = -2, ...,4 Corrado-Zivney statistics is defined as ([7]):
SAR„ = AR„ / S(AR{), where
SAR, t = -5 1 ,...,0 SARit / S(SARt) t = 1,..., 4, where
where n is the length of the pre-event window and rank(SAK'iaj denotes the rank of SAR'.m within the vector consisting of standardized abnormal returns from the pre-event window and SAtf . statistics is asymptotically normally distributed. The advantage o f the nonparametric Corrado-Zivney test is that is does not need any assumption o f abnormal returns normality.
Figure 1. Cross-sectional standard deviations of abnormal returns AR.t of WIG in the part of the pre-event window (t = -9, ..., -3) and in the event window (t = -2, ..., 4).
3. Empirical results
In the first step we examined the impact of each indicator separately ig noring interrelations between them as if the Employment Report contained in formation only about this one indicator. It means that for each indicator and for the same set o f abnormal returns data computed for all Employment Report announcements we define different events according to the value o f this indica tor. Detailed results o f the event study analysis for Unemployment Rate, Average Hourly Earnings and Average Workweek i.e. average abnormal returns in the event window and Corrado-Zinvey statistics Tcz are presented in Table 1 for all clusters. Nonfarm Payrolls equals consensus only in one case so only the impact of unexpected news about NP is examined and results o f the event study for NP are presented separately in Table 2. For each indicator the number of events in each cluster are also reported in Tables 1 and 2.
Table 1
Reaction of intraday WIG returns to unexpected news in the Employment Report of the U.S. Bureau of Labor Statistics in the period from January 2004 to November 2012
Below consensus In line with consensus Above consensus
AR(%) Tcz AR(%) Tcz AR(%) TC Z
Unemployment Rate
t 44 events 30 events 28 events
-2 0.007 0.24 -0.010 -0.68 0.012 1.40 -1 -0.004 -0.10 0.000 0.81 -0.004 -0.18 0 0.012 1.11 0.017 1.64 0.009 0.84 1 -0.016 -0.72 0.042 1.04 -0.091 -1.08 2 0.005 0.38 -0.016 -0.60 -0.018 -0.69 3 0.017 0.81 0.021 1.59 0.019 0.99 4 -0.012 -0.26 0.005 -0.09 -0.025 -0.35
Average Hourly Earnings
41 events 29 events 32 events
-2 0.006 0.42 -0.004 0.02 0.006 0.44 -1 -0.015 -1.19 -0.006 0.44 0.015 1.43 0 -0.008 0.43 0.031 2.58* 0.022 0.97 1 0.013 0.34 -0.047 -0.70 -0.037 -1.00 2 -0.032 -0.98 -0.020 -0.97 0.035 1.29 3 0.020 2.06“ -0.001 0.19 0.036 0.63 4 -0.010 -0.57 -0.005 -0.02 -0.016 -0.09 Average Workweek
35 events 45 events 22 events
-2 0.013 0.89 -0.001 -0.23 -0.003 0.34 -1 -0.007 -0.02 -0.016 -1.07 0.031 2.16” 0 -0.003 0.26 0.027 2.80* 0.009 0.43 1 -0.035 -0.60 -0.046 -0.87 0.057 0.43 2 -0.049 -1.47 0.035 1.40 -0.028 -0.94 3 0.041 1.68*** -0.012 -0.32 0.046 1.95*** 4 -0.001 0.38 -0.017 -0.57 -0.013 -0.88
*, **, *" - significant at 1%, 5% and 10% level respectively
In should be noted here, that new information released at t = 0 can be reflected in stock prices only for t greater than 0. Thus, the most important in the analysis of the reaction of WIG to the Employment Report is the value and significance of ARy
When the announced value of the Unemployment Rate is different from expecta tions stock prices decrease in the first five minutes after the announcement in both clusters “below consensus” and “above consensus”. On the other hand, when the unemployment rate is in line with expectations stock prices increase about 0.042%. It indicates that any unexpected information about unemployment is regarded by investors as bad news while UR in line with expectations is seen as good news. However, changes in abnormal returns induced by UR announcements are small when compared with returns standard deviation and hence become insignificant. These results confirm that in general the value of the unemployment rate in the U.S. does not impact stock prices. Similarly to UR, there is no visible impact of the announced value o f the Average Hourly Earnings. The absolute value of Tcx statistics in each cluster for t = 1 is smaller than any reasonable critical value. Opposite to UR, investors react differently depending on the value of AHE. The announcement below consensus is followed by positive average abnormal returns (0.013%) while AHE greater than forecast implies negative AR (-0.037%). It suggests that when AHE is smaller than expected it is seen as good news while AHE greater than expected is bad news. Corrado-Zivney test statistics is significant at least at 5% level only in two cases: when the announcement is below consensus (t = 3) and when the an nouncement is in line with consensus (t = 0). The significantly positive change in abnormal returns just before the announcement in line with forecasts is also observed when the Average Workweek is taken into account. AW shorter than ex pected is accompanied by negative average abnormal returns up to 10 minutes after the release. Then highly positive AR is observed and its significance is confirmed by Corrado-Zivney test. The reaction of returns to AW below consensus in the first 10 minutes after publication of the Employment Report means that the reduction of working time in the U.S. is bad news for investors. The opposite reaction is vis ible for AW above consensus when at first investors react positively. In this cluster increase in stock prices is confirmed by significant statistics for t = 3 It means that 15 minutes after unexpected news about Average Workweek highly positive investor reaction is observed on the WSE irrespective of the value of news.
The reaction of WIG returns to Nonfarm Payrolls is different from the reaction of any other indicator analyzed in this paper. When all the data are divided accord ing to unexpected news about Nfi then the strongest reaction to the Employment Report is observed. Moreover, the impact of news about NP is the easiest to inter pret. When NP is below consensus stock prices on the WSE decrease in average about -0.144% in the first five minutes after information release. This change is sig nificant at the 1% level (T = -3 96). After then there is no other significant changes. It means that NP smaller than expected is bad news to investors and their reaction is very strong, very fast and disappears after five minutes. The opposite reaction just after news release is observed when the announced value of NP is greater than
consensus (^47? L = 0.198%). Hence, NP above consensus is admittedly good news to the stock market as evidenced also by significantly positive ARy and by positive average abnormal returns in almost the entire event window.
Table 2
Reaction of intraday WIG returns to unexpected news about Nonfarm Payrolls in the period from January 2004 to November 2012
B elow con sen su s (6 4 even ts)
Above con sen su s (3 7 events) t AR(%) Tc z AR(%) TC Z ■2 0 .0 0 9 1.06 -0 .0 0 6 -0 .4 2 -1 -0 .0 0 4 0.29 0.001 0.18 0 0 .0 0 7 1.41 0.023 1.96*” 1 -0 .1 4 4 -3.96* 0 .1 9 8 4.07* 2 -0 .0 1 0 -0 .4 2 -0 .0 0 8 -0 .1 1 3 0 .0 1 0 0.38 0.035 2 .2 3 ” 4 -0 .0 2 0 -0 .8 0 0 .0 1 0 0 .7 7
- significant at 1%, 5% and 10% level respectively
Results in Table 1 and 2 indicate that unexpected changes in Nonfarm Payrolls are the major determinant of investor reaction to news about the U.S. labor mar ket included in the Employment Report. A significant reaction of WIG returns just after the Employment Report release is observed only when announcements are divided according to the value of NP Table 1 and 2, however, describe the reaction o f WIG returns to information contained separately in each of the indicators. To verify that stock prices change mainly according to news about Nonfarm Payrolls we repeat the above computations taking into account unexpected news about NP together with unexpected news about each of the other indicators. This allows us to determine the joint impact of pairs of the indicators. Because we examine the impact o f unexpected news, hence for each pair o f the indicators we divide data into four clusters: both indicators are above consensus, both are below consensus and one indicator is above consensus while the other is below (there are two such clusters). Detailed results of this event study are reported in Table 3 where we present mean abnormal returns AR in the event window together with statis tics. The small number o f events in some clusters could result in the low power of the preformed tests. However, results in Table 3 clearly indicate the importance of Nonfarm Payrolls. In each cluster the mean abnormal return for t = 1 is significant at least at 10% level and, what is even more important, the sign of AR depends only on the value o f unexpected news about NP Aft^s negative when NP is below
consensus and AR t is positive when NP is above consensus irrespective of the value of the other indicator. It confirms results from Table 1 and 2 that investors on the WSE react mainly to information contained in Nonfram Payrolls and the values of other indicators in the Employment Report are significantly less important.
Table 3
Reaction of intraday WIG returns to unexpected news about pairs o f indicators published in the Employment Report in the period from January 2004 to November 2012
Nonfarm Payrolls below consensus Nonfarm Payrolls above consensus
AR(%) Tc z AR(%) TC Z AR(%) TCZ A R (%) TCZ
UR below consensus UR above consensus UR below consensus UR above consensus
(31 events) (17 events) (17 events) (11 events)
- 2 0.009 0.55 0.021 1.50 0.004 -0 .1 9 - 0 . 0 0 1 0.38 - 1 -0 .0 0 4 -0 .0 8 -0 .0 0 7 -0 .1 5 -0 .0 0 3 0.11 0 . 0 0 1 -0 .1 0 0 0.009 1.33 0.003 0.20 0.021 0.33 0.019 1.20 1 -0 .1 7 6 -3.93* -0 .2 2 7 -2 .5 7 ” 0.338 3.10* 0.119 2.08** 2 -0 .0 0 8 0.12 -0 .0 2 4 -0 .8 2 0.046 0.94 -0 .0 0 8 -0 .1 8 3 0.003 -0 .0 9 0.017 0.72 0.040 1.10 0.022 0.61 4 -0 .0 2 0 -0 .1 7 -0 .0 4 5 -1 .0 7 0.006 -0 .2 7 0 .0 0 7 1.52
AHE below consensus AHE above consensus AHE below consensus AHE above consensus
(25 events) (21 events) (17 events) (11 events)
-2 0.010 0.38 0.012 0.88 0.003 0.39 -0 .0 0 6 -0 .4 6 - 1 -0 .0 2 2 -1 .1 4 0.018 1.33 -0 .0 0 4 -0 .3 6 0.009 0.59 0 -0 .0 2 4 -0 .4 5 0.022 0.97 0.016 1.75*” 0.023 0.31 1 -0 .1 0 5 Q 2 * ** -0 .1 6 3 -2 .4 7 ” 0.181 2.70* 0.205 1.65*** 2 -0 .0 3 5 -0 .7 7 0.022 0.48 -0 .0 2 1 -0 .2 9 0.060 2.10” 3 0.035 2.00” 0.025 -0 .2 6 - 0 . 0 1 1 0.48 0.056 1.35 4 -0 .0 2 4 -1 .0 8 -0 .0 3 7 -0 .7 2 0 . 0 1 1 0.16 0.023 1.63
AW below consensus AW above consensus AW below consensus AW above consensus
(26 events) (13 events) (8 events) (12 events)
-2 0.008 0.45 -0 .0 0 2 0.43 0.035 1.31 -0 .0 0 3 0.02 - 1 -0 .0 1 3 -0 .2 5 0.041 1.73” * 0.014 0.63 0.018 1.31 0 -0 .0 0 4 0.59 0 . 0 1 1 0.54 0 . 0 0 1 -0 .2 3 0.006 0.05 1 -0 .1 2 0 -2 .1 2 ” -0 .1 5 5 -2.72* 0.219 3.11* 0.312 3.03* 2 -0 .0 7 2 -2 .0 3 ” 0.004 0.02 0.033 1.41 -0 .0 6 6 -1 .6 3 3 0.029 1.02 0.039 0.82 0.066 1.24 0.055 1.87*** 4 0 . 0 0 0 0.35 -0 .0 2 7 -0 .5 8 -0 .0 0 2 0.15 0.004 -0 .6 6
4. Conclusions
The Employment Report published monthly by the U.S. Bureau o f Labor Statistics is one of the most significant U.S. macroeconomic announcements. In this paper we examined the impact of four macroeconomic indicators included in the Report (Unemployment Rate, Nonfarm Payrolls, Average Hourly Earnings and Average Workweek) on stock prices on the Warsaw Stock Exchange in the pe riod from January 2004 to November 2012. To analyze the impact o f unexpected information on 5-minute returns of WIG we apply the event study. The analysis shows that generally investors react to announcements about the U.S. labor mar ket in the first five minutes after information release. An unexpected increase in Average Hourly Earnings is seen by investors as bad news while increase is Average Workweek or Nonfarm Payrolls is good news. Reaction to unexpected information about the Unemployment Rate is negative. The comparison o f investor reaction to indicators under study shows that Nonfarm Payrolls is the most important in dicator i.e. changes in stock prices on WSE are in line with the value o f Nonfarm Payrolls irrespective to the value of other indicators in the Employment Report.
References
[1] Andersen T , Boilerslev T , Deutsche m ark - dollar volatility: intraday activity patterns, m acroeconom ic announcements, a nd longer run dependencies, “Journal of Finance” 1998, vol. 53, pp. 219-265.
[2] Andersen T , Bollerslev T , Diebold F., Vega C., Real-time p rice discovery in global stock, bond and foreign exchange markets, “Journal of International
Economics” 2007, vol. 73, pp· 251-277.
[3] Będowska-Sój ka B., Intraday CAC40, DAX a nd WIG20 returns ivhen the Am erican m acro news is announced, “Bank i Kredyt” 2010, vol. 41(2),
pp. 7-20
[4] Boyd J.H., Hu J., Jagannathan R., The Stock M arket’s Reaction to Unemployment News: Why Bad News Is Usually Good For Stock, “Journal of Finance” 2005, vol. 60, pp. 649-672.
[5 ] Campbell J.Y, Mei J . , Where do betas come from ? Asset p rice dynamics and the sources o f systematic risk, NBER Working Paper Series 1993, no. 4329-[6] Carnes WS., Slifer S.D., The Atlas o f Econom ic Indicators: A Visual Guide
to Market Forces a nd The Federal Reserve, Harper Collins Publishers, Washington, D.C.,
1991-[7] Corrado C.J., Event studies: A methodology revieiv, “Accounting and Finance” 2011, vol. 51, pp. 207-234.
[8] Corrado C.J., Truong C., Conducting event studies with Asia-Pacific security market data, “Pacific-Basin Finance Journal” 2008, vol. 16, pp. 493-521. [9] Corrado C.J., Zivney T.L., The specification and pow er o f the sign test in
event study hypothesis tests using daily stock returns, “Journal of Financial and Quantitative Analysis” 1992, vol. 27(3), pp. 465-478.
[ 10] Gurgul H., Suliga M., Wojtowicz T., Responses o f the Warsaw Stock Exchange to the U.S. M acroeconomic Data Announcem ent, “Managerial Economics” 2012, vol. 12, pp. 41-60.
[11] Harju K., Huissain S., Intraday Return and Volatility Spillovers Across International Equity Markets, “International Research Journal of Finance
and Economics” 2008, vol. 22, pp. 205-220.
[12] Nikkinen J., Sahlström E, Scheduled Domestic and US M acroeconomic Neivs and Stock Valuation in Europe, “Journal of Multinational Financial Management” 2004, vol. 14, pp. 201-245.